Personal Protective Equipment Monitoring in Construction Site Using Deep Neural Network

Abstract

Safety has always been of paramount importance in all industrial endeavors, particularly construction. It is not an ordinary office position, and precautions are necessary. Accidents and injuries are less likely to occur on a construction site when workers are well-equipped with safety gear. The cranium is the only organ entirely enclosed in bone in the human body. The significance of safeguarding the brain, a component of the body with a vital role in bodily function, is a natural law. Hard headwear and safety headgear serve as the first line of defense against head injuries, but only when worn correctly. Thus, it is reasonable to assert that safety Helmets reduce the risk of brain injury and save lives. This paper presents a system for real-time detection based on video streaming analysis and Deep Neural Network (DNN). A new method in convolutional neural network predicts whether or not employees are donning helmets correctly. The paradigm of edge computing in which the application for image analysis and classification is deployed on an embedded system directly connected to the camera. The proposed system is developed using a low-cost commercial embedded system, namely a Raspberry PI with an Intel Neural Compute Stick. The system was tested with various convolutional neural networks (CNNs) that had been pre-trained and was optimized to monitor the worker's headgear. In terms of classification performance and inference latency, CNNs were contrasted. Then, each CNN was deployed on the actual system and the system's throughput was measured by analyzing video frames per second.